CoRL paper: flow-matching VLAs can run 1-step Euler inference with no retraining
kaixhin · x · 2026-10-26
A CoRL 2026 paper shows that for flow-matching vision-language-action models, a single Euler inference step often matches multi-step inference for large, well-trained VLAs — no retraining or distillation needed.
The authors benchmark 1-step vs. reference multi-step success rates across LIBERO/LIBERO+, RoboTwin, CALVIN ABC→D, and a real-world robot suite, covering DIT, SmolVLA, X-VLA, FLOWER, π0.5, LAP, and XR-0. In most settings the gap stays within 3pp (≤3pp on 80% of LIBERO tasks, 74% on RoboTwin, 93% on CALVIN, 67% real-world), implying nearly free inference speedups at deployment.
Related event: One-Step Euler Inference Matches Multi-Step Flow Matching in VLA Models(2 posts)→
More from Embodied
- Inside Dexory's full-stack warehouse robotics factory: 8m scanning towers and in-house manufacturing — lukas_m_ziegler · 2026-10-26
- M5Stack StickS3 firmware adds remote live view for Ricoh GR III/IIIx/IV cameras — OpenAIDevs · 2026-10-26
- Fourier's GR-3 humanoid climbs six floors autonomously, descends stairs untethered — chris_j_paxton · 2026-10-26
- FocusVLA routes robot camera views by predicted interaction intent across task stages — iscas-cloud · 2026-10-26
- Pandora M1: first prototype of a physical device for distributed GPU monitoring — tarantulae · 2026-10-26
- $150 modular Linux handheld Mecha Comet opens preorders — pocket SSH and Claude Code included — Paimaamu · 2026-10-26